tobotiq - robotics bin picking
Post on 28-Jan-2016
32 Views
Preview:
DESCRIPTION
TRANSCRIPT
BIN PICKING APPLICATIONS
AND TECHNOLOGIES
2
TABLE OF CONTENTS
INTRODUCTION .................................................................................................................................. 3
TYPES OF MATERIAL HANDLING ......................................................................................................... 3
WHOLE BIN PICKING PROCESS ........................................................................................................... 4
VISION SYSTEM: HARDWARE .................................................................................................... 4
VISION SYSTEM: SOFTWARE ..................................................................................................... 5
END EFFECTORS ........................................................................................................................ 6
BASIC METRICS TO MEASURE THE QUALITY OF A SOLUTION ............................................................ 7
PACKAGED SYSTEMS FOR BIN PICKING .............................................................................................. 8
CONCLUSION .................................................................................................................................... 11
3
INTRODUCTION
Either for machine tending or part transfer, material handling is worth automating since it is
monotonous and a non-value added job. Robots are already used in manufacturing to do these
tasks. However, there still room for improvement. The first step of any manufacturing application
is related to bin picking, since the robot has to pick up the part prior to doing its job. As often
mentioned, automated bin picking applications are the “Holy Grail” of robotics since they have
many challenges to overcome.
There are different types of bin picking with different levels of difficulty. The main features that
define the complexity of bin picking applications are:
1. The geometry of the part.
2. The severity of randomness of parts in the bin.
This document will provide you with information about what you should know and consider in
order to automate your bin picking applications. First, let’s take a look at the different kinds of
material handling. Then, we will talk about the whole bin picking process from cameras and
software to selection and gripping. After reading this document, you should be able to define your
bin picking process and have all the necessary information to find the best packaged system to suit
your needs.
TYPES OF MATERIAL HANDLING
STRUCTURED MATERIAL HANDLING
In this type of manipulation, the rigid part that needs to be gripped has a predetermined location
and orientation easily reachable by a robot and its end effector. It either needs jigs or it arrives at
your manufacturing plant in a structured bin.
SEMI-STRUCTURED MATERIAL HANDLING
Here the part is constrained on a planar surface with no assigned location or orientation. This is
defined as a 2 degrees-of-freedom (DOF) pose. There is no overlapping of parts. This kind of
handling is easily done with the help of a vision system under controlled lighting.
BIN PICKING
Bin picking consists of a single type of object randomly
piled in a box. The system needs to resolve the pose of
each object with 6 DOF. A lot of overlapping parts may be
encountered. Here again, actual vision technologies have
proven that they can solve this kind of puzzle, but
controlled lightning is still needed.
4
WHOLE BIN PICKING PROCESS Bin picking requires the synchronization of three components:
1. Vision Systems: the hardware component
Basically, a robot has to locate and grasp objects that have been placed randomly in a
container. To do so, a vision system is used to take pictures of the bin.
2. Vision Systems: the software processing
This part is all about the data analysis scanned or detected by the hardware component.
The images scanned are processed by the software to recognize a part in the bin to be
picked up. Then, the software needs to talk with the robot controller to plan the best path
to pick the selected part up avoiding any collision.
3. The end effectors
Finally, the robot moves to grasp the object by using its end effector and brings it to the
next manufacturing cell.
1. VISION SYSTEM: HARDWARE
Different vision technologies now exist on the market. Vision systems don’t only include the
cameras; the computation of the information received by them is also part of the system. Once
images are taken, the software needs to process the data in order to recognize the part to pick.
Their methods can differ from one manufacturer to another, but the vision system itself can be
categorized.
2D ROBOT VISION
This kind of vision system can be used for identification and position
recognition of a part on a 2D plane. It usually can detect the X, Y variation
and the rotation of the part around the Z axis. There is usually only one
camera located at a constant distance from the part.
2.5D ROBOT VISION
The 2.5D vision can also detect the part’s height using a single camera. The
camera takes different pictures of the part or it can also use a time-of-flight
laser to determine its height. This kind of laser will calculate the time needed
for the laser light to hit the part and come back. In order to take its different
poses, the camera of this vision system needs to move. The information
received from the images can define the position of the part in X, Y, Z and its rotation around the
vertical axis. This kind of vision is a lean system equivalent to the ability of a 3D camera and is
often used for de-palletizing.
5
3D ROBOT VISION
3D vision can detect variations in X, Y and Z direction. Moreover, it can detect
rotation around the three axes giving 6 Degrees-Of-Freedom (DOFs) part
location. It is the preferred technology for bin picking applications. 3D
systems can use either one moving camera or two with a shape-matching
(model). A TOF laser with surface-matching can also be used (point cloud).
The camera distances can change.
FIXED-CAMERA VS END-OF-ARM MOUNT
The way the cameras are mounted can present some advantages or disadvantages.
A fixed-Camera will allow the robot to keep going with its work while the picture is taken and
processed. However, the camera needs to be far from the moving arm to avoid any interference.
So, the pictures are often less accurate and not always in the right orientation. To capture more
than one angle of the part, another camera is usually needed.
When the camera is mounted on the robot arm, the accuracy of the picture is improved.
Moreover, the orientation can be adjusted to get the best images possible. One camera can be
used to get pictures of the part from different angles. However, the camera needs to have high-
flex cabling because it is moving with the robot arm. If the camera is in harm's way, then it might
be damaged if a collision occurs. Another thing to consider is that the arm needs to stop in order
to take a picture, so the cycle time can be longer if it is mounted on the robot arm.
2. VISION SYSTEM: SOFTWARE
The computational aspect of the vision system consists of two tasks: object recognition and path
planning.
a. The Object Recognition
Object recognition uses the local geometric features of the part to be able to accurately calculate
the position and orientation of the object. Moreover, objects that contain a lot of unique features
can be detected from a partial view. The vision system, usually a 2D or 3D camera, takes pictures
of the bin, then processes the images in order to recognize a part in the bin. Different methods are
used such as comparisons with CAD models, 3D point cloud, geometric pattern matching (GPM),
etc. Object recognition not only gives information on the detected object, but also on the
undetected ones. This is called environmental information. By using this information, collisions can
be avoided while calculating the path.
6
b. The Path Planning
This leads us to the path planning role of the vision system. When an object’s position and
orientation is determined, the path of the robot for reaching this object is calculated. As stated
before, collisions have to be avoided, so the vision system helps by providing information on the
other parts in the bin, as well as for the localization of the bin’s walls. This is a safety feature that
all bin picking systems should be using.
END EFFECTORS
When all the calculations are done, the robot arm can move
and pick the piece detected. However, to grab this object,
the right end effector has to be chosen. And here again
there are some challenges to consider:
Complexity of robot gripper design
Specific robot grippers can be specially designed for
the object to be handled in the bin. But, since the
parts are randomly located in the bin with different
orientations, designing the right gripper can be a
complex task.
End effector flexibility
Also designing a specific device brings flexibility problems, because different grippers need
to be used for different pieces. Furthermore, what if the bin contains various types of
objects?
Having the right type of end effector according to the parts to be handled
Common grippers can be used to perform bin picking. However each one has its pros and
cons. For example, suction cups may be effective for small and flat parts having a regular
finish. However, if you have an irregular part with a porous finish, things could be different.
Same thing with pneumatic grippers: having parts in a structured or semi-structured
7
environment with enough space between them could be a good solution. However if the
end effector has to reach into a confined space and deal with overlapping parts, pneumatic
grippers may not be the perfect fit due to the difficulty of doing partial opening and
closing.
Precision required once the part is picked
Depending on the applications and the process, picking the part might be only one part of
the challenge. The accuracy needed for the next step in the production process can also
have an impact on the choice of end effector. Indeed if the part has to be inserted in a CNC
machine, put on a conveyor or brought to a particular assembly position the precision
required will have an impact on what you need in terms of handling equipment.
Industrial bin picking applications seem to represent a good fit with flexible electric grippers.
These kinds of robot gripper provide more flexibility by, among other things, performing partial
opening and closing to reach tight areas.
BASIC METRICS TO MEASURE THE QUALITY OF A SOLUTION
The potential metrics that could be used to evaluate a solution for bin picking applications are the
following:
Speed
Which is defined as the time required to pick an object in the bin or the number of objects which
can be picked in a given amount of time.
Efficiency
Which is a mixture of time efficiency (time required to locate the object vs cycle time), grasping
quality and percentage of success.
Accuracy
Which is the measurement of error in object recognition and part pose estimation.
Since there are a lot of factors influencing these metrics, the best way to evaluate a solution is to
test it. A bin picking system is not only about the cameras or the end effector; it is the combination
of every component in the cell. This is why it is difficult to determine the quality of a solution
considering the components individually.
8
PACKAGED SYSTEMS FOR BIN PICKING
MAJOR BRANDS
FANUC
Fanuc offers a complete solution for bin picking applications.
All the hardware and software is 100% from Fanuc. This
solution is called iRVision 3DL and it is integrated with the R-
30iA Controller. The system can work with either one or more
cameras. No additional hardware is necessary such as a PC.
You just have to plug the cameras to the controller. There is a
light exposure mode to compensate for light fluctuation. The
software has a lot of options such as inspection module, GPM
tool locator to detect a previously trained image pattern, BLOB
tool to detect the part even if it has irregularities, curve tool locator and more. Moreover, the
iRCalibration Suite offers calibration modules for various systems such as vision shift, TCP set,
vision frame set and more. For more information, visit their website:
http://www.fanucrobotics.se/en/products/vision-systems/irvision
YASKAWA – MOTOMAN
Motoman has partnered with world-leading vision product
providers to offer a packaged solution for bin picking
applications. Their new software MotoSight 3D offers real
3D vision to their robot arms. This vision software supports
6 DOFs with one, two or three cameras. If the single
camera option is chosen, it has to be mounted on the
robot. If multiple cameras are used, they can either be
mounted on the robot or fixed to another location. This
gives better flexibility and can accommodate irregular parts. It can be integrated with off-the-shelf
cameras and it is supported by their DX100 and NX100 controllers. The software will compare
what it sees to a standard CAD file of your part. In term of performance, the whole system can
handle 6 to 10 part per minute. For more information, here is their website:
http://www.motoman.com/products/vision/#sthash.zx9TKcVS.OYIdBnIP.dpbs
9
ABB
The vision system integrated for ABB robots is named TrueView.
It is compatible with the IRC5 controller and all IRB robot arms.
The patented single camera 3D technology can locate a part
with 6 DOFs. The camera is mounted on the robot arm and is
supplied with super flex cables for a long service life. The set-up
is fast as well as the calibration. The camera used can be an
analog Sony CCD, JAI or Baumer both of which are GigE digital
cameras. The system has 2 red LED light bars to improve the
quality of the pictures taken. The vision processing time is about
1 to 3 seconds. For more information, follow this link to get the
datasheet:
http://www05.abb.com/global/scot/scot241.nsf/veritydispla
y/599b2f4b48524fe8c1257b130057a117/$file/TrueView%
205.12%20revB.pdf
INDEPENDENTS
SHAPESCAN
3D SHAPEscan is a solution, developed by ISRA Vision, to precisely locate the position of a part
which is randomly arranged in a bin. This system includes 2 cameras and special laser lighting. The
system can be installed in a fixed position, as well as on a robot arm. It can be adjust for any size of
container by adjusting the distance between the two cameras. The 3D sensor scans the bin and
the software develops a so called 3D point cloud. With this newly developed evaluation method,
SHAPEscan can identify basic geometric shapes in the bin. It assures robust recognition of a part’s
geometry. The depth is determined by laser projection. The stereo view, allowed by the two
cameras, makes it capable to see in 3D just as a human does. For more information on this
solution, get the datasheet via this link: http://www.isravision.com/en/products/robot-vision/3D-
shape-scan
10
SCAPE TECHNOLOGIES
Scape Technologies offers a standardized solution for any bin
picking application. Based in Denmark, this company enables robot
arms to see in order to grip unsorted objects in a bin. The
installation is very simple: Install the software on a standard PC,
mount the Scape Tool Unit on the robot, use the standard Scape
software, upload the files of your parts to the PC and, by making
some adjustments, the system is ready to start running. The Bin
Picking Suite offers a range of recognition and gripping options.
This standardized solution also offers an optimal return on
investment. To get more information, visit their website:
http://www.scapetechnologies.com/
These packaged systems are just a few of the solutions available on the market. Many different
independent vision integrators are providing great solutions for bin picking applications.
Therefore, to make the right choice, it is not only important to ask about their products, but also
about their services, since vision technology is still about testing and adjustments.
11
CONCLUSION
For the past two decades, robot manufacturers have been working hard to put in place efficient
robotic bin picking systems. It still is one of the most difficult applications for robotics. But now
with the technical progress of 3D vision systems, computation and flexible end-effectors, this
application seems to be a reality for today's industrial day-to-day production line. Many solutions
have been proven to work with success, so it is now time to convince the industrial world that
solutions for bin picking are available.
ABOUT ROBOTIQ
Robotiq designs and manufactures flexible robotic tools for agile automation. We aim to give to
industrial manufacturing – from large businesses to SMEs – robotic tools solutions to optimize
automated processes. We work with robot manufacturers, system integrators and end-users to
enable new applications and improve productivity. Contact us for more information at
info@robotiq.com. Or visit our website at www.robotiq.com
12
© 2008-2013 Robotiq, Inc. 966 chemin Olivier, suite 325, St-Nicolas, Québec, Canada, G7A 2N1. All Rights Reserved.
top related